Improved Electricity Consumption Predictions with AI-Powered Daily Demand Forecasting

Economists have pinned Energy as the Achilles heel of the Turkish economy. The rapid expansion of incomes and industry has logically led to a rise in the demand for energy in Türkiye—a nation with very limited supplies of fossil fuels.  

This has caused Türkiye to become heavily dependent on imports—which account for roughly 70 percent of energy consumption. Türkiye also has:

• No nuclear power plants
• A higher share of coal usage than the EU
• A higher proportion of Hydro-power than in the rest of Europe—with a lower proportion currently derived from other renewable energy sources 

Hence, Turkish energy suppliers have a small margin for error. Planning departments responsible for energy trading are the first line of defense and must adequately handle any number of events, including:

• Equipment failure in turbines, generators, and transformers
• Sudden decreases in production due to fuel shortages or natural disasters
• Sudden increases in production due to unexpected spikes in demand or changes in contractual obligations
• Fluctuations in energy prices and changes in government regulations
• Cyber attacks

Therefore, as the Turkish economy continues to grow and the country becomes more urbanized—accurate Electricity Consumption forecasts are vital for electric distribution companies so they can deliver the most consistent day ahead and intra-day forecasts.

AutoML Turns Hours of Manual Planning Into Seconds of Successful Daily Forecast

Founded as the first company of Zorlu Energy Group in 1993, Zorlu Enerji is one of the most prestigious leaders in the Turkish energy sector. The wholesale division of Zorlu Energy includes customers who have chosen Zorlu Energy as their electricity distributor. Their portfolio includes both :

• Large customers (hotels and shopping centers)
• Relatively small consumers (individual customers who get their electricity from the Zorlu family) 

Depending on the amount of electricity consumed, the unit price of electricity—for an individual residence, workshop, factory, or larger factory—fluctuates and the system pays more as it is consumed. 

This can lead to situations where it’s less advantageous to buy electricity from a wholesaler—and customers can start to buy electricity from a residential tariff. By being able to properly estimate imbalances in portfolios, Zorlu can have more flexibility to understand how much power they will need to draw each hour during the next month—and help serve and retain customers. 

When customers give notice about entering or exiting the portfolio in the first five days of a given month—they can be included or leave the portfolio as of the first day of the following month.

The most crucial step is estimating hourly electricity consumption for the portfolio for the following day.

For example, if Zorlu makes daily estimates themselves, it will take one hour for half an hour a day. Making manual predictions allows them to see the historical data more clearly. It’s relatively easy, for example, to predict the amount of electricity consumption for a hotel—or a manufacturer undergoing maintenance. 

By using ML Studio, Zorlu Energy can train separate models for different hourly clocks—taking advantage of hourly weather forecasts and historical usage data. 

The result? By using ML Studio, Zorlu Energy has been able to:

• Give accurate, ML-based predictions—alleviating the need for manual predictions
• Automate manual predictions—turning each hour of predictions into just a few seconds
• Incorporate their business knowledge into their models

Would you like to see how ML Studio can help your organization solve demand forecasting issues or other real-world problems—with a robust and ready-to-use End-to-End AI platform that doesn’t go off budget? 

Get in touch for a personal demo.